Saturday, December 22, 2012
the First principle of data analytics is to avoid making mistakes (continued)
Mistakes may happen when data or processes flow from one system to another. Sometimes, mistakes are caused by the flaws in the analytic methodologies themselves. One of the popular mistakes is to think in terms of absolute numbers, not ratios. For example, analysts see a few hundreds of fraudulent credit card transactions happening at a particular online retailer website and conclude that the retailer is very risky. This may not be true. If the retailer has tens of millions of normal transactions within the same period, the ratio of fraudulent transactions is very low. We have seen in number of occasions that people use number of frauds,instead of fraud rate, as the measurement of risk. As a result, they create fraud detection rules that generate a lot of false alarms.